| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Full paper |
Health Physics, Gartnavel Royal Hospital, West House, Glasgow G12 0XH, UK
Correspondence: Dr Colin John Martin, Head of Health Physics, Department of Clinical Physics and Bio-Engineering, Gartnavel Royal Hospital, West House, 1055 Great Western Road, Glasgow G12 0XH, UK. E-mail: colin.martin{at}northglasgow.scot.nhs.uk
 |
Abstract |
|---|
 Top Abstract IntroductionMethods and materialsResultsDiscussionConclusionReferences |
|---|
|
Introduction |
|---|
|
TopAbstractIntroductionMethods and materialsResultsDiscussionConclusionReferences |
|---|
E is not a quantity that can be measured. It can only be derived by computation, so that models and simulations are required to estimate the doses to individual tissues. In addition, E does not relate directly to the relative risk of harm in an individual, as there are known differences with age and sex. The application of E in its present form was recommended by the International Commission on Radiological Protection (ICRP) [3], which stated that it was intended for use in radiation protection, including the assessment of risks in general terms. However, E has been applied extensively to medical exposures, often to specific individuals of known gender and age. This method of application of E is inappropriate, and its primary use for medical exposures should be restricted to comparing the health detriment (i.e. the stochastic radiation risks) to a reference patient for different types of medical examination. Such relative comparisons can be used in both the generic justification and the optimization of medical exposures, but should not be used to predict absolute risk levels. Values have been derived for a variety of diagnostic procedures in radiology and nuclear medicine in order to provide a relative index of harm that can be considered in justification of medical exposures.
For medical exposures, conversion coefficients have been derived that allow values for E to be calculated from measurable dose quantities, such as entrance surface dose (ESD) or dose–area product (DAP) for radiology examinations [6, 7], and administered activity for nuclear medicine procedures [8, 9]. The coefficients have been established from computer simulations for the exposure of anthropomorphic phantoms. The coefficients are quoted to two or three significant figures, but the uncertainties in these and in the tissue weighting factors are seldom considered. The ICRP do not quote uncertainties in the tissue weighting factors with regard to assessing doses from prospective exposures for specification of radiological protection measures, in order to avoid the addition of unnecessary complexity to the calculation. However, in the extension of the application of E to assessment of doses from medical exposures, the underlying approximation of the method must be remembered. More reliance is often placed on E than the data on which it has been established can support, and this could lead to decisions being made on the basis of unsound evidence. In this article, estimates are made of uncertainties in the calculation of E for a reference patient for a selection of common medical exposures. The degree of confidence that can be placed on E as a relative indicator of risk of health detriment for medical procedures is considered and how this relates to numerical predictions of risk. The role that E should play in the assessment of medical exposures is reviewed and recommendations made about its application.
|
Methods and materials |
|---|
|
TopAbstractIntroductionMethods and materialsResultsDiscussionConclusionReferences |
|---|
|
|
CRt/Ct) reflects the accuracy with which tissue volumes and positions for a reference patient and the radiation interactions within them can be represented. It has been assumed that this approximates to a Gaussian distribution for the selection of tissues sensitive to radiation. For the systematic component of the error, which will depend on the method, the fractional error is assumed to be similar for each tissue and is represented by CS/CS. The tissue weighting factors have been assigned so that the sum of all the factors is 1.0. Thus, positive errors in some will be balanced by negative errors in others. The fractional uncertainty in Fsum due to the individual tissue weighting factors (wt/wt) and the random error component in the organ dose conversion coefficients can be expressed as:
|
|
DM/DM) are included, the fractional error in E is given by:
|
|
, and these have been used in calculations in this paper. The error in DM will be determined by the accuracy of the experimental measurement and the dose meter calibration and, for nuclear medicine, this will be replaced by the uncertainty in administered activity. Standard errors considered to be representative of standard practice are given in the first section of Table 1.
|
For nuclear medicine procedures, data from biokinetic models, which are used to calculate the number of radioactive disintegrations occurring within specific tissues and body regions, are combined with data from dosimetric models, from which the deposition of energy in relevant target organs of a reference patient is derived. The uncertainties in the biological quantities have often not been well characterized [23, 24]. Radionuclides are assumed to be distributed uniformly throughout the source organs. Variations in the estimated cumulated activity in target organs arise from uncertainties in uptake, distribution and retention, and will give rise to a systematic component in the uncertainty. The random uncertainties arise from the size of the individual tissues and the distances between source organs and irradiated tissues. Calculations have shown that estimates of absorbed doses in different organs do not generally deviate by more than a factor of three from actual absorbed doses, and the deviation is less for pharmaceuticals labelled with short-lived radionuclides such as technetium-99m [23]. Experimental validation has indicated agreement within the range 20–60% [9]. Random uncertainties of ±40% for each tissue have been assumed with a systematic component of ±30% for the standard error in E for a reference patient.
Tissue weighting factors
The tissue weighting factors are based on probability coefficients for individual tissues or organs for an aggregated health detriment [3]. It is difficult to account for uncertainties in choice of model to fit the epidemiological data linking cancer incidence to organ dose from which the weighting factors were derived so, for the purpose of the initial assessment, it is assumed that the linear no threshold and relative risk dose–effect models used are correct. The uncertainties in the model used for extrapolation of the data will be considered in the discussion. The weighting factors have been grossly rounded to facilitate ease of use and because the system was acknowledged to be an approximation. Differences between the weighting factors and the corresponding normalized aggregated detriment for a population of all ages and both sexes on which they are based [3] are portrayed as a percentage difference for each tissue in Figure 1. These values have been used to represent the uncertainties in tissue weighting factors. They have been combined with values for organ dose coefficients [6, 9, 25] in calculations of Fsum/Fsum in order to allow an assessment to be made of the uncertainty in E as a quantity linked to relative harm.
|
). The weighting factor for the breasts has been increased because of a rise in the health detriment based on recent data from the LSS cohort, in which breast cancer accounted for 18% of the radiation-associated solid cancers averaged over males and females, compared with 11% from the previous assessment. The proposed increase in weighting factor has been set larger than the rise in mean risk of incidence in order to avoid underprotection of women. The reduction in weighting factor for the gonads is based upon recent judgements on the risk of hereditary effects [27], which include all classes of hereditary effects up to the second post-irradiation generation, rather than the genetic risks at a theoretical equilibrium in the population on which the current factor is based [3]. The ratios of E derived using the proposed 2007 weighting factors and the current values have been calculated.
|
|
Results |
|---|
|
TopAbstractIntroductionMethods and materialsResultsDiscussionConclusionReferences |
|---|
and Figure 1 are between 20% and 50% (Figure 2). These uncertainties apply to E for a reference patient of mass 70 kg with doses and risks averaged between a male and female for each tissue. If the procedures in Figure 2 are regarded as a representative selection of medical examinations, the uncertainty in the calculation of E for a medical exposure for a reference patient is about ±40% for an 80–90% confidence limit. In order to give an indication of the variations resulting from refinement of the tissue weighting factors, the ratios of values for E calculated using the proposed 2007 weighting factors and the current values [28] for a selection of diagnostic medical examinations are shown in Figure 3. The revision of the weighting factors changes E by less than 20% for the majority of exposures of the trunk, but there are larger changes for some examinations involving fewer organs, such as those in dental radiology.
|
|
) [3, 26, 30, 31]. Adjustment factors to the risk for different ages could be applied in principle to take care of these differences. However, the variations shown in Figure 4 may be significantly different for the partial body exposures involved in diagnostic medical examinations, so this approach is not appropriate.
|
|
Discussion |
|---|
|
TopAbstractIntroductionMethods and materialsResultsDiscussionConclusionReferences |
|---|
). For the purposes of assessing relative risk to different tissues, based on current knowledge, the uncertainty in E for a reference patient is about ±40% for an 80–90% confidence limit (Figure 2). However, this is not the whole story. The assumption was made for the purpose of the calculation of uncertainties that the aggregated probability coefficients represented an accurate reflection of the risk of health detriment [3]. They provide the best estimate based on evidence available at the time but, because they are derived from a study of one population group made using an extrapolation from much higher doses, there is considerable uncertainty in the associated risks. In order to decide how much reliance can be placed on E as an index of harm, it is necessary to consider the evidence on which it is based, namely the Japanese LSS. There are uncertainties in the data from the cancer diagnoses, the individual dose reconstructions and statistical variations in the genetic make-up and size of individuals among the population. Additional uncertainties arise from the type of risk model used to translate the results derived from one population to another with different baseline rates of cancer incidence, the method of extrapolation from results for high to low dose levels, and the models used to predict cancer risks as a function of age at exposure and time since exposure. Considering first the uncertainty in the risk model, excess risk associated with radiation exposure for each type of cancer can be described in terms of the absolute risk or a relative risk related to the natural incidence of cancer in the population. The absolute risk model assumes constant excess risks after a latent period, while the relative risk model projections are based on the background age-specific death rates for the different cancers in the particular population. Natural incidences of cancer can vary significantly between different populations. For example, comparing the age-standardized incidence rates per 100 000 per year for stomach and breast cancer, these are 31 and 34, respectively, in Japan and 3 and 90, respectively, in the United States [32, 33]. Nor do these rates remain static, so that, over the last 30 years in the UK, the age-standardized incidence of breast cancer has increased from 75 to 117 per 100 000 per year, although the mortality rate has declined, while the incidence of stomach cancer has declined from 15 to 7 per 100 000 per year [34]. There is epidemiological evidence that the absolute risk model is more realistic for breast cancer [35, 36], while the relative risk model may be more appropriate for stomach cancer [37, 38]. However, there is insufficient evidence to allow alternatives to be ruled out in these cases, while there is little relevant information for other cancers. The relative risk model for a general population of all ages in the UK or US based primarily on data from the Japanese LSS tends to predict more deaths from cancer than the absolute risk model. Although there are significant differences for cancer in individual tissues, when data for all tissues are summed, the difference in cancer deaths is less than a factor of two. As more of the deaths associated with the relative risk model occur later in life, the aggregated health detriment for the two models, which makes allowance for years of life lost, will be smaller.
There has been considerable debate about the extrapolation of the epidemiological results to low doses. Most studies conclude that there is no convincing evidence of a higher incidence of cancer among individuals who received doses less than 100–150 mSv in any of the epidemiological studies [26, 33], and these doses are much higher than those encountered in either occupational exposures or diagnostic medical ones. Some groups argue that the adaptive responses of cells enhance repair of DNA damage, which will provide protection at low doses [39], but this is disputed by others [40, 41]. A recent assessment of epidemiological study data from the LSS that are now available concluded that there is good evidence for an increased risk of incidence [4] and mortality [5] for solid tumour in humans for acute doses of X-rays down to about 10–50 mSv [42]. This may be due partly to the presence of subpopulations at greater risk on account of age [26, 31], genetic status [29] or other factors. Studies of radiation workers and of populations exposed to higher levels of background radiation have not demonstrated convincing evidence of higher rates of cancer for chronic exposures, although borderline levels of significance have been found for incidence of both solid tumour and leukaemia in some studies [26, 43]. However, such studies are problematic because of the size of the study population required to provide the statistical power to demonstrate an effect, and the difficulty in accounting for other differences between populations, which might produce a bias that masks the effects of radiation. If there is a dose threshold, this is unlikely to exceed 60 mGy [4].
The linear no threshold approach, in which a linear extrapolation of data on excess cancer risk vs equivalent dose to the organ at risk is made [3], provides the most appropriate practical model for judging radiation effects. A dose and dose rate effectiveness factor (DDREF) is introduced, by which risk estimates derived from Japanese LSS data are reduced, in order to allow for the tissue protection and repair mechanisms that are known to protect against damage. However, the magnitude of the DDREF, which has been set at 2, has a large uncertainty. Factors of between 2 and 5 have been found from studies in animals, and the possibility of values between 1 and 5 is acknowledged [3, 26, 45]. The uncertainties in the model used to extrapolate to other populations and the DDREF, and the possibility that there may be a dose threshold for the effects, means that actual risks to a reference patient at low doses could be a factor of 2.5–3 higher or lower [45–47]. When the other uncertainties discussed are included, the risks of health detriment for individuals could be more than five times higher or lower.
As E is based on the risk of health detriment, it can be used with probability coefficients to estimate the excess lifetime risk of developing fatal cancer following a radiation exposure [3]. Risks of cancer derived from these coefficients are often quoted in dose reports and radiological protection publications. However, the uncertainties discussed mean that risk estimates for a reference patient could be higher or lower than the accepted value by a factor of 3, and that, for an individual, they may be higher or lower by a factor of 5. The use of numerical estimates of risk gives a false impression of certainty and is not appropriate. Risks can only be grouped within a range covering a factor of about 10. Therefore, it is recommended that general terms are used for describing risk based on E. A variety of terms have been proposed [39, 48, 49]. Those proposed by the UK Department of Health [48], which cover broad ranges of risk, are probably the most appropriate for general use in discussions of risks with patients (e.g. low is between 1 in 1000 and 1 in 10 000, and very low is between 1 in 10 000 and 1 in 100 000). Based on the current nominal probability coefficient for fatal cancer averaged over the whole population of 5x10–2 Sv–1 [3], these would apply to dose ranges of 2–20 mSv and 0.2–2 mSv, respectively. However, the most convenient dose ranges to specify are 0.1–1 mSv, 1–10 mSv, etc., both in numerical terms and for the range of examinations that they cover, and, as the uncertainty in the risk is in any event large, it would be inappropriate to make the nominal risks the foundation of the risk scale. The effects from low doses of ionizing radiation will not occur until many years after the exposure, therefore the length of life lost by exposed individuals will be lower than for the risks of immediate death that apply to most other causes. It is therefore proposed that the term "very low" risk is used for the dose range 1–10 mSv, which equates to risks of dying from cancer over a lifetime of between 1 in 2000 to 1 in 20 000, as calculated using the linear extrapolation dose–effect model. The scale and terms proposed for risks in the range covered by diagnostic medical exposures, all of which are below the level at which there is definitive evidence of radiation effects in humans, are given in Table 3. For any E less than 0.1 mSv which is over 100 times less than the dose for which there is any evidence of a link to cancer induction, and similar to the dose received from natural background radiation over a period of 2 weeks, the risk is considered negligible.
|
). The most significant changes in the revised tissue weighting factors are a higher factor for the breasts and a smaller one for the gonads (Table 2). Other minor changes in tissue weighting that have been proposed are additional factors of 0.01 for the brain and salivary glands, while the remainder, which is a collection of organs regarded as potentially at risk, has been expanded to include more organs, and the associated remainder weighting factor increased. Examinations for which the change in E has been more than 20% are ones of the pelvis, where the gonads are one of the most highly exposed organs, and dental exposures such as dental orthopantomographic examinations, because the salivary glands, which have been allocated their own weighting factors, lie within the primary beam [28]. The current system for calculating E includes a proviso that, if the dose to one of the remainder organs is higher than that to any other radiosensitive tissue, then half of the remainder organ weighting factor would be allocated to this tissue, called the "splitting rule" [3]. As a result, E for head CT, in which the brain is the most exposed organ, is lower with the new system. Effective dose evolves as more knowledge is gained, but changes in relative weighting factors are unlikely to have a major effect on E for exposures of the trunk, but will have a greater influence for regions of the body where only a limited number of organs are exposed.
Application of effective dose
So where does this leave us? There are uncertainties in the quantity E, but it is the only quantity available that provides a dose related to the risk of health detriment. Effective dose has been employed in many applications for which it was never intended, and much greater accuracy has been attributed to it than is justified. It is not intended to provide an individual specific dose, but the dose to a reference person. The uncertainty in E as a relative value for comparison of risk to a reference patient for different types of exposure is about ±40% for an 80–90% confidence limit, but the uncertainties in absolute numerical risks of cancer for an individual are far greater. Because of the uncertainty in the organ dose conversion coefficients and tissue weighting factors, and the approximate nature of the method used to evaluate the risk, it is only appropriate to use numeric assessments of E as an expression of approximate risk-related doses for reference patients. It is therefore reasonable that E should be used in order to assess relative risks to a reference patient, which are then described in general terms such as low, very low, minimal and negligible. However, E does not relate to an individual, but to a reference hermaphrodite patient for whom risks have been assessed based on the average for a whole population. When dealing with risks to individual patients from radiation exposure, the best indicator is obtained by estimating the mean doses to all radiosensitive organs in the exposed individual and combining these with the latest age-, sex- and organ-specific risk coefficients for radiation-induced stochastic effects available in ICRP reports and the published literature. In this context, the variation in risk with age is particularly important (Figure 4) and must be taken into account. A suggested approach for the application of E and risk assessment to medical applications is summarized in the following:
), or for choosing a risk category relating to a reference patient in describing the health detriment from a small unintended overexposure. |
Conclusion |
|---|
|
TopAbstractIntroductionMethods and materialsResultsDiscussionConclusionReferences |
|---|
Received for publication September 4, 2006. Revision received October 25, 2006. Accepted for publication October 30, 2006.
|
References |
|---|
|
TopAbstractIntroductionMethods and materialsResultsDiscussionConclusionReferences |
|---|
pez M, Morant JJ, Calzado A. Monte Carlo calculations of radiation dose in CT examinations using phantom and patient tomographic models. Radiat Prot Dosim 2005;114:364–8.This article has been cited by other articles:
 |
|
 |
|
  E. Angel, N. Yaghmai, C. M. Jude, J. J. DeMarco, C. H. Cagnon, J. G. Goldin, C. H. McCollough, A. N. Primak, D. D. Cody, D. M. Stevens, et al. Dose to Radiosensitive Organs During Routine Chest CT: Effects of Tube Current Modulation Am. J. Roentgenol., November 1, 2009; 193(5): 1340 - 1345. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 F. A. Mettler Jr, M. Bhargavan, K. Faulkner, D. B. Gilley, J. E. Gray, G. S. Ibbott, J. A. Lipoti, M. Mahesh, J. L. McCrohan, M. G. Stabin, et al. Radiologic and Nuclear Medicine Studies in the United States and Worldwide: Frequency, Radiation Dose, and Comparison with Other Radiation Sources--1950-2007 Radiology, November 1, 2009; 253(2): 520 - 531. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 R. Fazel, H. M. Krumholz, Y. Wang, J. S. Ross, J. Chen, H. H. Ting, N. D. Shah, K. Nasir, A. J. Einstein, and B. K. Nallamothu Exposure to Low-Dose Ionizing Radiation from Medical Imaging Procedures N. Engl. J. Med., August 27, 2009; 361(9): 849 - 857. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J. E COLANGELO, J. JOHNSTON, J. B KILLION, and D. L. WRIGHT Radiation Biology and Protection Radiol. Technol., May 1, 2009; 80(5): 421 - 441. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 G Dietze, J D Harrison, and H G Menzel Effective dose: a flawed concept that could and should be replaced. Comments on a paper by D J Brenner (Br J Radiol 2008;81:521-3) Br. J. Radiol., April 1, 2009; 82(976): 348 - 351. [Full Text] [PDF]
|
|
|
|
|
|
A. Sodickson, P. F. Baeyens, K. P. Andriole, L. M. Prevedello, R. D. Nawfel, R. Hanson, and R. Khorasani Recurrent CT, Cumulative Radiation Exposure, and Associated Radiation-induced Cancer Risks from CT of Adults Radiology, April 1, 2009; 251(1): 175 - 184. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. R. Mayo and J. A. Leipsic Radiation Dose in Cardiac CT Am. J. Roentgenol., March 1, 2009; 192(3): 646 - 653. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 T. C. Gerber, J. J. Carr, A. E. Arai, R. L. Dixon, V. A. Ferrari, A. S. Gomes, G. V. Heller, C. H. McCollough, M. F. McNitt-Gray, F. A. Mettler, et al. Ionizing Radiation in Cardiac Imaging: A Science Advisory From the American Heart Association Committee on Cardiac Imaging of the Council on Clinical Cardiology and Committee on Cardiovascular Imaging and Intervention of the Council on Cardiovascular Radiology and Intervention Circulation, February 24, 2009; 119(7): 1056 - 1065. [Full Text] [PDF]
|
|
|
|
|
|
P. C. Shrimpton, B. F. Wall, T. T. Yoshizumi, L. M. Hurwitz, and P. C. Goodman Effective Dose and Dose-Length Product in CT Radiology, February 1, 2009; 250(2): 604 - 605. [Full Text] [PDF]
|
|
|
|
|
|
J A ROBERTS, N A DRAGE, J DAVIES, and D W THOMAS Effective dose from cone beam CT examinations in dentistry Br. J. Radiol., January 1, 2009; 82(973): 35 - 40. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 A. J Einstein Radiation risk from coronary artery disease imaging: how do different diagnostic tests compare? Heart, December 1, 2008; 94(12): 1519 - 1521. [Full Text] [PDF]
|
|
|
|
 |
|
 F. R. Verdun, F. Bochud, F. Gundinchet, A. Aroua, P. Schnyder, and R. Meuli Quality Initiatives Radiation Risk: What You Should Know to Tell Your Patient1 RadioGraphics, November 1, 2008; 28(7): 1807 - 1816. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
F. R. Verdun, F. Bochud, F. Gudinchet, A. Aroua, P. Schnyder, and R. Meuli Radiation Risk: What You Should Know to Tell Your Patient RadioGraphics, September 24, 2008; (2008) 287085042. [Abstract] [Full Text]
|
|
|
|
 |
|
 J. B. Ludlow, L. E. Davies-Ludlow, and S. C. White Patient Risk Related to Common Dental Radiographic Examinations: The Impact of 2007 International Commission on Radiological Protection Recommendations Regarding Dose Calculation J Am Dent Assoc, September 1, 2008; 139(9): 1237 - 1243. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 K. Smans, L. Struelens, M. Smet, H. Bosmans, and F. Vanhavere PATIENT DOSE IN NEONATAL UNITS Radiat Prot Dosimetry, August 30, 2008; (2008) ncn237v1. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
K. Smans, L. Struelens, M.-T. Hoornaert, F. Bleeser, N. Buls, D. Berus, P. Clerinx, F. Malchair, F. Vanhavere, and H. Bosmans A study of the correlation between dose area product and effective dose in vascular radiology Radiat Prot Dosimetry, July 1, 2008; 130(3): 300 - 308. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
F. A. Mettler Jr, W. Huda, T. T. Yoshizumi, and M. Mahesh Effective Doses in Radiology and Diagnostic Nuclear Medicine: A Catalog Radiology, July 1, 2008; 248(1): 254 - 263. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
D J BRENNER Effective dose: a flawed concept that could and should be replaced Br. J. Radiol., July 1, 2008; 81(967): 521 - 523. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
C. J. Martin RADIATION DOSIMETRY FOR DIAGNOSTIC MEDICAL EXPOSURES Radiat Prot Dosimetry, March 28, 2008; (2008) ncm495v1. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
C. J. Martin The application of effective dose to medical exposures Radiat Prot Dosimetry, January 1, 2008; 128(1): 1 - 4. [Full Text] [PDF]
|
|
|
|
|
|
P P Dendy Radiation risks in interventional radiology Br. J. Radiol., January 1, 2008; 81(961): 1 - 7. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| BJR | DMFR | IMAGING | ALL BIR JOURNALS |
